VLSI Low Power Design Analysis Modeling Strategies

Complex simulation models are an important tool in the decision-making process. To increase the ability to efficiently explore broader subspaces of these models, this dissertation presents a series of experimental designs for searching over as many as 22 variables in as few as 129 runs. These unique designs blend an orthogonal Latin hypercube with uniform patterns to generate designs that are both space-filling and nearly orthogonal. Multiple metrics are utilized to assess the quality of candidate designs and select the best one. When more than the required number of runs is available, the designs can be permuted and appended to create additional design points, which increase the design's orthogonality and space-filling qualities. Two alternative surfaces are investigated using the designs.